Keyboard shortcuts

Press ← or → to navigate between chapters

Press S or / to search in the book

Press ? to show this help

Press Esc to hide this help

RTL / BiDi QA Corpus (Chat, Markers, UI, Subtitles, Closed Captions)

Canonical test-string appendix for RTL/BiDi/shaping/font-fallback/layout-direction validation across runtime UI, D059 communication, and D038 localization/subtitle/closed-caption tooling.

This page is an implementation/testing artifact, not a gameplay feature design.

Purpose

Use this corpus to validate that IC’s RTL/BiDi support is correct beyond glyph coverage:

  • text shaping (Arabic joins)
  • bidirectional ordering (RTL + LTR + numerals + punctuation)
  • wrap/truncation/clipping behavior
  • font fallback behavior (theme primary font + fallback backbone)
  • D059 sanitization split (legitimate RTL preserved, spoofing controls handled)
  • replay/moderation parity for normalized chat and marker labels

This corpus supports the execution-overlay clusters:

  • M6.UX.RTL_BIDI_GAME_UI_BASELINE
  • M7.UX.D059_RTL_CHAT_MARKER_TEXT_SAFETY
  • M7.UX.D059_BEACONS_MARKERS_LABELS
  • M9.SDK.RTL_BASIC_EDITOR_UI_LAYOUT
  • M10.SDK.RTL_BIDI_LOCALIZATION_WORKBENCH_PREVIEW
  • M11.PLAT.BROWSER_MOBILE_POLISH

How To Use This Corpus

Runtime UI / D059 Chat & Markers (M6 / M7)

  • Render strings in:
    • chat log
    • chat input preview
    • ping label / tactical marker label
    • replay viewer communication timeline
    • moderation/review UI excerpts
  • Validate:
    • same normalized bytes and visible result in all of the above
    • marker semantics remain icon/type-first (labels additive only)
    • no color-only dependence

D038 Localization Workbench (M10)

  • Load corpus entries as preview fixtures for:
    • briefing/debrief text
    • subtitles
    • closed captions (speaker labels, SFX captions)
    • radar comm captions
    • mission objective labels
    • D065 tutorial hints / anchor overlays
  • Validate:
    • line wrap/truncation
    • clipping/baseline alignment across fallback fonts
    • layout-direction preview (LTR / RTL) behavior

Platform Regression (M11)

  • Re-run a subset of this corpus on:
    • Desktop
    • Browser
    • Steam Deck
    • Mobile (where applicable)
  • Compare screenshots/log captures for layout drift.

Test Categories

A. Pure RTL (Chat / Labels / UI)

Use these to validate shaping and baseline RTL ordering without mixed-script complexity.

IDStringLanguage/ScriptPrimary Checks
RTL-A1هدفArabicArabic shaping/joins; no clipping in marker labels
RTL-A2إمداداتArabicCombined forms/diacritics spacing; fallback glyph coverage
RTL-H1גשרHebrewCorrect RTL order; marker-label width handling
RTL-H2חילוץHebrewBaseline alignment + wrap in narrow UI labels

B. Mixed RTL + LTR + Numerals (High-Value)

These are the most important real-world communication cases for D059 and D070.

IDStringIntended ContextPrimary Checks
MIX-1LZ-בMarker labelMixed-script token order, punctuation placement
MIX-2CAS 2 هدفTeam chat / marker noteNumeral placement + spacing under BiDi
MIX-3גשר A-2Objective / markerLatin suffix + numerals remain readable
MIX-4Bravo 3 חילוץChat / quick noteLTR word + numeral + RTL tail ordering
MIX-5יעד: Power Plant 2Objective text / subtitleRTL punctuation + LTR noun phrase
MIX-6טניה: Move now!Closed caption (speaker label)RTL speaker label + LTR dialogue text ordering

C. Punctuation / Wrap / Truncation Stress

Use these to catch line-wrap and clipping bugs that a simple glyph test misses.

IDStringContextPrimary Checks
WRAP-1מטרה: השמידו את הגשר הצפוני לפני הגעת התגבורתObjective panelMulti-word wrap in RTL layout; punctuation placement
WRAP-2هدف المرحلة: تعطيل الدفاعات ثم التحرك إلى نقطة الاستخراجBriefing/subtitleArabic wrap + shaping under multi-line width
TRUNC-1LZ-ב צפון-מערבMarker labelEllipsis/truncation in bounded marker UI; no clipped glyph tails
TRUNC-2CAS יעד-2 עכשיוSmall HUD calloutShort-width truncation preserves intent/icon semantics
WRAP-3‏[انفجار بعيد] تحركوا إلى نقطة الإخلاء فوراًClosed caption (SFX + speech)Mixed caption prefixes/brackets + Arabic wrap/shaping

D. D059 Marker Label Bounds (Byte + Rendered Width)

These are tactical labels that should stay short. They validate D059’s dual bounds (normalized bytes + rendered width).

IDStringExpected Result ClassNotes
LBL-1AAAcceptBaseline ASCII tactical label
LBL-2גשרAcceptPure RTL short label
LBL-3LZ-בAcceptMixed-script short label
LBL-4CAS 2AcceptLTR+numerals tactical label
LBL-5יעד-חילוץ-צפוןTruncate or reject per width ruleValidate deterministic width-based handling
LBL-6هدف-استخراج-الشمالTruncate or reject per width ruleArabic shaping + width bound behavior

Rule reminder: Behavior (accept / truncate / reject) may vary by UI surface policy, but it must be documented, deterministic, and replay-safe.

E. Font Fallback / Coverage Validation (Theme Primary + Fallback Backbone)

Use these when the active theme primary font is likely missing Arabic/Hebrew glyphs.

IDStringPrimary Checks
FB-1Mission: חילוץLatin primary + Hebrew fallback glyph run selection
FB-2CAS → هدفLatin + symbol + Arabic fallback; spacing and baseline alignment
FB-3יעד 2 / LZ-BMixed-script + numerals + punctuation across fallback runs
FB-4توجيهات الفريقPure Arabic fallback shaping and clipping

Must validate:

  • no tofu/missing-glyph boxes in supported locale/script path
  • no clipped ascenders/descenders after fallback
  • no line-height jumps that break HUD/chat readability

F. D059 Sanitization Regression Vectors (Escaped / Visible Form)

These are sanitization harness inputs. Represent dangerous characters in escaped form in tests; do not rely on visually invisible raw literals in docs.

Goals

  • preserve legitimate RTL content
  • block or strip spoofing/invisible abuse per D059 policy
  • keep normalization deterministic and replay-safe
IDInput (escaped notation)Example IntentExpected Validation Focus
SAN-1\"ABC\\u202E123\"BiDi override spoof attemptDangerous control handled (strip/reject/warn per policy); visible result deterministic
SAN-2\"LZ\\u200B-ב\"Zero-width insertion abuseInvisible-char abuse handling without breaking visible text semantics
SAN-3\"גשר\\u2066A\\u2069\"Directionality isolate/control experimentPolicy-consistent handling + replay parity
SAN-4\"هدف\\u034F\"Combining/invisible abuseCombining-abuse clamp behavior deterministic

Policy note: This corpus does not redefine the allowed/disallowed Unicode policy. D059 remains canonical. These vectors exist to prevent regressions and ensure moderation/replay tools show the same normalized text users saw in-match.

G. Replay / Moderation Parity Checks

For a selected subset (MIX-2, LBL-3, SAN-1, SAN-2):

  1. Submit via chat or marker label in a live/local test.
  2. Capture:
    • chat log display
    • marker label display
    • replay communication timeline
    • moderation/review queue snippet (if available in test harness)
  3. Verify:
    • normalized text bytes are identical across surfaces
    • visible result is consistent (modulo intentional styling differences)
    • no hidden characters reappear in replay/review tooling

H. Layout Direction Preview Fixtures (D038 / D065)

Use these strings to verify LTR vs RTL layout preview without changing system locale:

IDStringSurfacePrimary Checks
DIR-1התחל משימהButton / action rowAlignment, padding, icon mirroring policy
DIR-2هدف المرحلةObjective cardCard title alignment in RTL layout profile
DIR-3Press V / לחץ VD065 tutorial hintMixed-script instructional prompt + icon spacing
DIR-4CAS Target / هدف CASD070 typed support marker tooltipTooltip wrap + semantic icon retention

I. Closed Caption (CC) Specific Fixtures

Use these to validate CC formatting details that differ from plain subtitles (speaker labels, SFX cues, bracketed annotations, stacked captions).

IDStringSurfacePrimary Checks
CC-1טניה: אני בפנים.Cutscene/dialogue closed captionRTL speaker label + RTL dialogue shaping/order
CC-2Tanya: אני בפנים.Cutscene/dialogue closed captionLTR speaker label + RTL dialogue ordering
CC-3[אזעקה] כוחות אויב מתקרביםSFX + speech captionBracketed SFX cue placement and wrap in RTL
CC-4[انفجار] Tanya, move!SFX + mixed-script dialogueArabic SFX cue + LTR speaker/dialogue ordering
CC-5דיווח מכ״ם: CAS 2 מוכןRadar comm captionAcronyms/numerals inside RTL caption remain readable

CC-specific checks:

  • Speaker labels and SFX annotations must remain readable under BiDi and truncation rules.
  • Caption line breaks must preserve meaning when labels/SFX prefixes are present.
  • If the UI uses separate styling for speaker names/SFX cues, styling must not break shaping or reorder text incorrectly.

If time is limited, run these first:

  • RTL-A1
  • RTL-H1
  • MIX-2
  • LBL-3
  • FB-2
  • SAN-1
  • DIR-3
  • CC-2

This set catches the most common false positives:

  • “glyphs render but BiDi is wrong”
  • “chat works but markers break”
  • “fallback renders but clips”
  • “sanitization blocks legitimate RTL”
  • “subtitle works but closed-caption labels/SFX prefixes reorder incorrectly”

Maintenance Rules

  • Add new corpus strings when a real bug/regression is found.
  • Prefer stable IDs over renaming existing cases (keeps test history diff-friendly).
  • If a string is changed, note why in the linked test/bug/ticket.
  • Keep this page implementation-oriented; policy changes still belong in:
    • src/02-ARCHITECTURE.md
    • src/decisions/09g-interaction.md
    • src/decisions/09f-tools.md